Computable error bounds for an optimization problem with parallelepiped constraint

It can be shown that interval arithmetic can be used to locate the Kuhn-Tucker point of a nonlinear programming problem where its constraint is a parallelepiped with sides parallel to the coordinate axes.